How neuromorphic processors are reshaping autonomous combat | Electronics360
Move From Laboratory to Battlefield
By Stephen L. Pendergast, Contributing Correspondent | San Diego
Bottom Line Up Front
What Is a Neuromorphic Processor?
Conventional computers — from the laptop on a pilot's desk to the mission computer aboard an F-35 — operate on principles codified by John von Neumann in 1945: a central processor fetches instructions and data from memory, executes them, and writes results back, continuously and regardless of whether anything meaningful is happening in the outside world. That architecture is extraordinarily powerful for deterministic tasks but profligate with energy when applied to the irregular, sparse, real-world sensor streams that define battlefield autonomy.
Neuromorphic processors take a fundamentally different approach. Inspired by the structure of the human brain, they consist of networks of artificial neurons and synapses that communicate by firing discrete electrical pulses — "spikes" — only when sensor inputs cross a meaningful threshold. Between events, circuits remain dormant. The result, known as a Spiking Neural Network (SNN), processes data with power consumption measured in milliwatts rather than the watts or tens of watts demanded by graphics processing units and conventional AI accelerators.
The concept originated with Carver Mead's work at Caltech in the late 1980s, but scalable silicon implementations have only recently become manufacturable. Today's leading commercial neuromorphic processors include Intel's Loihi 2, fabricated on a 4-nanometer process and containing over one million programmable neurons and 120 million synapses, and BrainChip Holdings' Akida AKD1000, the first fully digital, event-based neuromorphic processor to reach commercial production. IBM's TrueNorth architecture, evaluated by the U.S. Army, completes the principal Western tier. Beyond these, Australia's Western Sydney University unveiled the "Deep South" system in 2024, built on 100 FPGA boards capable of simulating 100 billion neurons at 220 trillion spikes per second — a scale the project's developers claim matches human-brain-level processing.
The key operational advantage is event-driven selectivity. A neuromorphic chip monitoring an airspace radar feed does not process every sample in a 250-MHz data stream; it reacts only to anomalous returns — the micro-Doppler signature of a spinning rotor, the distinctive radar cross-section flutter of a low-observable drone — and fires only then. For a sensor-laden autonomous system operating on a lithium-polymer cell, that selectivity extends mission endurance dramatically.
— Sean Hehir, CEO, BrainChip Holdings
The AFRL Radar Contract: Neuromorphic Meets Micro-Doppler
The clearest indicator that neuromorphic processing has moved from academic paper to defense contract is a $1.8 million award announced April 1, 2025, from the Air Force Research Laboratory's Small Business Innovation Research program. Under contract topic AF242-D015, titled "Mapping Complex Sensor Signal Processing Algorithms onto Neuromorphic Chips," BrainChip and Raytheon — an RTX business — are collaborating to demonstrate neuromorphic radar processing focused on micro-Doppler signature analysis.
Micro-Doppler refers to the modulations imposed on a radar return by the sub-target motions of rotating or vibrating components — rotor blades, walking limbs, vehicle treads — embedded within a larger Doppler shift. Extracting that signature with conventional fast Fourier transform pipelines requires sustained computational throughput and commensurate power. Mapping the algorithms onto Akida's event-driven hardware, if the contract's goals are validated, could embed sophisticated discrimination capability into thermally and power-constrained weapon systems: missiles, small drones, and drone-defense interceptors where the computing budget is measured in watts, not kilowatts.
"The initiative builds on prior successful demonstrations of radar processing algorithms on our Akida neuromorphic hardware," BrainChip CMO Steven Brightfield said at the time of the announcement. "These demonstrations were independently validated by the RTX team and by ISL, who each reported significant performance benefits." The AFRL contract is a Phase II SBIR award — meaning at least one earlier technical milestone has already been met.
Independent academic work is converging on similar findings. A 2025 preprint in Neuromorphic Computing and Engineering from European researchers demonstrated spiking neural resonators applied to frequency-modulated continuous-wave (FMCW) radar, achieving state-of-the-art range-and-angle estimation while transmitting only 0.02 percent of the data volume required by a conventional Fourier-transform pipeline. The researchers noted that their neuron model "significantly reduces processing latency compared to traditional frequency analysis algorithms" because it emits results continuously rather than waiting to accumulate a full data frame.
Lockheed Martin's ForwardEdge Collaboration
Raytheon is not the only defense prime embedding neuromorphic AI in its silicon roadmap. On March 17, 2026, BrainChip announced a strategic collaboration with ForwardEdge ASIC — a wholly owned subsidiary of Lockheed Martin specializing in advanced application-specific integrated circuit architecture and microelectronics development. The partnership's stated objective is to embed BrainChip's Akida neuromorphic AI engines directly into ForwardEdge's custom silicon, enabling what the companies describe as "cognitive processing closer to the sensor," reducing data movement, lowering power consumption, and enabling autonomous operation in complex operational environments.
"BrainChip's Akida architecture is a strong complement to our ASIC and RF platforms, allowing us to integrate dedicated AI acceleration directly into the silicon," said Bill Jenkins, Chief Revenue Officer at ForwardEdge ASIC. The collaboration targets aerospace and defense systems requiring low-latency, onboard autonomous decision-making — categories that encompass fighter aircraft sensor fusion, missile seekers, space-based surveillance payloads, and counter-drone systems.
Defense industry analysts at Yahoo Finance observe that the partnership positions Lockheed Martin to differentiate its defense electronics proposals against rivals including Northrop Grumman, RTX, and BAE Systems as programs of record increasingly specify onboard AI processing requirements that conventional CPUs and GPUs cannot meet within SWaP envelopes.
| Platform / Organization | Neuromorphic Chip | Application | Status / Date |
|---|---|---|---|
| Raytheon / BrainChip (AFRL) | Akida AKD1000 | Micro-Doppler radar signal processing | Active contract, Apr 2025 |
| Lockheed Martin ForwardEdge ASIC | Akida (custom ASIC integration) | Onboard cognitive sensing, RF/EW systems | Collaboration announced Mar 2026 |
| Raytheon AVC / Universities | Akida AKD1000 | Autonomous UAV/UGV navigation & collaborative landing | Competition ongoing, 2025–2026 |
| Intel / DARPA programs | Loihi 2 | Autonomous vehicle navigation, battlefield sensor fusion | Multiple DARPA MTO programs |
| IBM TrueNorth / U.S. Army | TrueNorth | Evaluation for drone swarm detection | Research evaluation |
| Everspin (CHEETA Program / Purdue–DoD) | MRAM neural accelerator | Low-thermal-signature onboard processing | Production demo targeted 2026 |
| SynSense Speck / European researchers | Speck neuromorphic device | Battery-powered drone detection (event camera + SNN) | Published 2025 |
Operation Touchdown: Building the Talent Pipeline
Hardware contracts alone do not generate a defense neuromorphic ecosystem; the workforce to design, integrate, and sustain these systems must be cultivated in parallel. Raytheon's answer is the Autonomous Vehicle Competition, now in its 2025–2026 cycle under the theme "Operation Touchdown." BrainChip, announced as the official technology sponsor in February 2026, is providing its Akida AKD1000 processors at cost to undergraduate engineering teams across four U.S. geographic regions: South (University of Texas at Arlington), East Coast (George Mason University), West Coast (Santa Barbara City College), and Puerto Rico.
The competition's signature technical challenge encapsulates the core neuromorphic value proposition in miniature: teams must program a UAV to autonomously detect, track, and land on a moving UGV — no human intervention, no remote guidance, no cloud uplink. The task demands real-time computer vision, sensor fusion, and cooperative control, all within the power envelope of a small-format airframe. Events at UT Arlington and George Mason were completed in April 2026; Santa Barbara follows in June.
"Rather than traditional classroom exercises, these competitions present students with actual engineering constraints," BrainChip CEO Sean Hehir noted. "Students who participate develop a worldview in which defense work is rewarding and meaningful, and they acquire the applied skills needed for these industries." Raytheon's competition lead, Jesse Lee, characterized it as an effort to "push the boundaries of what university students can achieve in autonomous systems."
The Market and the Money
The commercial and defense market context gives operational urgency to these engineering decisions. According to Market Intelo's March 2026 research report, the global neuromorphic computing for defense sector was valued at $1.8 billion in 2025 and is projected to reach approximately $9.7 billion by 2034 at a compound annual growth rate of 20.5 percent. North America dominated the 2025 market with a 42.3-percent revenue share — roughly $762 million — underpinned by federal funding from DARPA's Microsystems Technology Office, the Air Force Research Laboratory, and congressional appropriations that Market Intelo calculates exceeded $1.1 billion for brain-inspired computing research between 2022 and 2025 under successive National Defense Authorization Acts. The DoD committed an estimated $285 million in neuromorphic defense contracts in the 2025–2026 period through DARPA's MTO alone.
The broader drone market context amplifies the stakes. Market Intelo projects global drone market revenues will exceed $35 billion by 2030, with AI-driven computer vision now considered a baseline requirement across virtually all UAV platforms under development. Neuromorphic chips' ability to deliver that vision capability at a fraction of the power of GPU-based inference makes them structurally attractive to any program manager whose platform is battery-limited.
BrainChip, whose Akida product line is the only commercially available, fully digital, event-based neuromorphic AI processor, closed a $25 million capital raise in December 2025 ahead of CES to fund commercialization of its AKD1500 module — designed for rugged industrial PC deployments — and its "Akida Pico" ultralow-power neural processing unit core, now available for remote evaluation via FPGA cloud. Grand View Research projects the overall neuromorphic computing market will reach $20.27 billion by 2030, growing at 19.9 percent annually from 2024.
Intel's Hala Point and the Scale Question
While BrainChip occupies the tactical, SWaP-constrained edge of the market, Intel's neuromorphic program operates at a different scale. In 2024 Intel delivered "Hala Point," the world's largest neuromorphic computer, built on an array of Loihi 2 chips. The system supports more than one billion neurons and 128 billion synapses and consumes 2,600 watts — significant by neuromorphic standards but orders of magnitude below what a conventional supercomputer would require to simulate equivalent network complexity. Intel's Mike Davies, director of neuromorphic computing at Intel Labs, described Hala Point as demonstrating "that neuromorphic architectures can achieve breakthroughs in efficiency and scale" applicable to real-time robotics and autonomous navigation research.
Loihi 2 itself, manufactured on Intel's 4-nanometer process and featuring over one million programmable neurons, has been deployed in DARPA-funded autonomous vehicle programs and is described by Market Intelo as "commercially the most advanced neuromorphic processor deployed in defense research settings." Its open-source Lava software framework lowers the barrier for AI developers to port existing spiking neural network models to neuromorphic hardware, addressing what has historically been the technology's principal adoption obstacle.
Electronic Warfare and Counter-Drone Applications
Beyond radar signal processing, neuromorphic chips are attracting interest for electronic warfare and counter-drone missions precisely because those domains generate the sparse, irregular sensor data streams that event-driven architectures handle most efficiently. A 2025 paper from European researchers published through arXiv demonstrated a fully neuromorphic drone detection system combining an event camera (a sensor that outputs pixel-level brightness changes rather than full video frames) with an SNN running on SynSense's Speck chip. The system achieved detection accuracy competitive with GPU-based reference implementations while consuming "several orders of magnitude" less energy — allowing, the authors calculated, more than a year of continuous operation on battery power.
Parallax Advanced Research, whose Dr. Steven Harbour leads AI hardware research, has argued that third-generation AI techniques including SNNs address a fundamental brittleness in conventional AI: the inability to generalize to unexpected inputs. In electronic warfare — an environment defined by adversarial adaptation and novel jamming waveforms — that adaptability advantage could be decisive. Parallax has explored integration of neuromorphic processors into existing electronic countermeasure pods carried by both Air Force and Navy strike packages.
The U.S. Army's SBIR office has separately funded research into ultra-fast neuromorphic drone-swarm detection using artificial antiferromagnetic neurons — a nanoscale magnetic hardware approach capable of identification tasks in sub-nanosecond timeframes with power consumption below 1 picojoule per synaptic operation, extending neuromorphic principles into GHz and potentially THz frequency ranges potentially relevant to future high-frequency radar and electronic attack systems.
Meanwhile, Everspin Technologies is developing MRAM-based neural accelerators under the CHEETA program, a DoD initiative led by Purdue University, in which computation occurs entirely within the magnetic memory array. That architecture eliminates the energy cost of moving data between processor and memory — historically consuming up to 90 percent of the energy budget in AI tasks — and produces a thermal signature low enough to satisfy military hardware requirements for non-detectability in contested environments.
Doctrine, Ethics, and the Accountability Question
The doctrinal implications of deploying systems that can "see, think, and react without a tether" have not escaped the notice of the National Defense University. In a 2025 commentary published through NDU's Digital Commons, Dr. James Giordano of the Center for Disruptive Technology and Future Warfare argued that neuromorphic AI's ability to engage in what he terms "causal decision processing" — reasoning about cause and effect rather than merely pattern-matching — distinguishes it qualitatively from prior generations of military AI and demands correspondingly more sophisticated oversight frameworks.
Giordano concludes that "integrating such systems into defense operations must be guided by clear oversight, doctrine, and ethical frameworks to ensure effectiveness, safety, and stability in future military engagements." That position aligns with the broader scholarly consensus emerging in 2025–2026 around autonomous weapons governance. A July 2025 paper in the journal Defence Studies contends that autonomous weapons systems "fundamentally undermine moral accountability in war" when deployed without structures that preserve human agency in lethal decision-making — a concern that becomes more pointed as neuromorphic chips approach the processing speed and adaptability needed to operate effectively in denied-communications environments where human-in-the-loop oversight is architecturally unavailable.
The tension is structural: the very feature that makes neuromorphic processors operationally attractive — the ability to act autonomously on sparse sensor data without uplink dependency — is precisely what complicates the accountability chain that international humanitarian law requires. DoD's existing autonomous weapons policy (DoD Directive 3000.09) mandates "appropriate levels of human judgment over the use of force," but that directive was written before event-driven, milliwatt-scale AI accelerators capable of battlefield discrimination tasks existed in production silicon.
Outlook
The velocity of recent program announcements — an AFRL radar contract in April 2025, a Lockheed Martin subsidiary collaboration in March 2026, a multi-university neuromorphic UAV competition running concurrently — suggests that neuromorphic processing has crossed the threshold from technology readiness to program integration. The outstanding questions are not whether these chips will appear in fielded defense systems, but at what rate, in which mission sets, and under what command-authority structures.
What the Ukraine and Middle East conflicts have established beyond reasonable dispute is that the SWaP math for autonomous systems is ruthlessly unforgiving. A drone that can think longer between charges, discriminate targets with less processing overhead, and operate without a communications tether is a fundamentally more capable weapon than one that cannot. Neuromorphic processors may not be the only answer to that equation — but as of mid-2026, they are the most credible one moving from laboratory to flight line.
Verified Sources & Formal Citations
-
Brown, Peter. "How Neuromorphic Processors Are Reshaping Autonomous Combat." Electronics360 / GlobalSpec, May 2026.
https://electronics360.globalspec.com/article/23688/how-neuromorphic-processors-are-reshaping-autonomous-combat -
BrainChip Holdings Ltd. "BrainChip Partners with RTX's Raytheon for AFRL Radar Contract." Press release, April 1, 2025.
https://brainchip.com/brainchip-partners-with-rtxs-raytheon-for-afrl-radar-contract/ -
BrainChip Holdings Ltd. "BrainChip Named Official Technology
Sponsor for Raytheon's 'Operation Touchdown' Autonomous Vehicle
Competition." Business Wire / BrainChip.com, February 26, 2026.
https://brainchip.com/brainchip-named-official-technology-sponsor-for-raytheons-operation-touchdown-autonomous-vehicle-competition/ -
BrainChip Holdings Ltd. "ForwardEdge ASIC Selects BrainChip's
Neuromorphic Computing for Future ASICs." Press release, March 17, 2026.
https://brainchip.com/forwardedge-asic-selects-brainchips-neuromorphic-computing-for-future-asics/ -
BrainChip Holdings Ltd. "BrainChip Announces $25 Million (USD)
Funding Ahead of CES to Power Next-Gen Edge AI." Business Wire, December
10, 2025.
https://www.businesswire.com/news/home/20251210514718/en/BrainChip-Announces-$25-Million-USD-Funding-Ahead-of-CES-to-Power-Next-Gen-Edge-AI -
Intelligence Community News. "BrainChip Collaborates with RTX's Raytheon on AFRL Contract." April 2, 2025.
https://intelligencecommunitynews.com/brainchip-collaborates-with-rtxs-raytheon-on-afrl-contract/ -
Market Intelo. "Neuromorphic Computing for Defense Market Research Report 2034." March 10, 2026.
https://marketintelo.com/report/neuromorphic-computing-for-defense-market -
Market Intelo. "Neuromorphic Computing Chips Market Research Report 2034." March 9, 2026.
https://marketintelo.com/report/neuromorphic-computing-chips-market -
Giordano, James. "Beyond Mechanistic Control: Causal Decision
Processing in Neuromorphic Military Artificial Intelligence." Center for
Disruptive Technology and Future Warfare, Institute for National
Strategic Studies, National Defense University, 2025. Digital Commons @ NDU, Commentary No. 1.
https://digitalcommons.ndu.edu/commentary/1/ -
Intel Corporation. "Neuromorphic Computing and Engineering with AI — Loihi 2." Intel Labs, 2025.
https://www.intel.com/content/www/us/en/research/neuromorphic-computing.html -
Semiconductor Insight. "Intel Launches World's Largest
Neuromorphic Supercomputer as Brain-Inspired AI Goes Mainstream." July
14, 2025.
https://semiconductorinsight.com/blog/intel-launches-worlds-largest-neuromorphic-supercomputer-as-brain-inspired-ai-goes-mainstream/ -
Parallax Advanced Research. "Parallax Advanced Research and the
Future of Neuromorphic Artificial Intelligence in Electronic Warfare."
2025.
https://parallaxresearch.org/news/blog/parallax-advanced-research-and-future-neuromorphic-artificial-intelligence-electronic -
[Authors redacted for blind review]. "Drone Detection Using a
Low-Power Neuromorphic Virtual Tripwire." arXiv preprint
arXiv:2509.12997, 2025.
https://arxiv.org/abs/2509.12997 -
[Authors]. "Range and Angle Estimation with Spiking Neural Resonators for FMCW Radar." Submitted to Neuromorphic Computing and Engineering, arXiv preprint arXiv:2503.00898, 2025.
https://arxiv.org/abs/2503.00898 -
U.S. Army SBIR/STTR Program. "Drone Swarm Detection Using
Artificial Intelligence Based on Ultrafast Neural Networks." SBIR Topic,
Army Research Office.
https://armysbir.army.mil/topics/drone-swarm-detection-ai-based-ultrafast-neural-networks/ -
Yahoo Finance / Investor Analysis. "Lockheed Martin Bets On
Neuromorphic AI Chips For Future Defense Edge." March 20, 2026.
https://finance.yahoo.com/sectors/technology/articles/lockheed-martin-bets-neuromorphic-ai-120356206.html -
Uplatz Blog. "The Convergence of Spiking Neural Networks and
Neuromorphic Hardware: An In-Depth Analysis of Intel's Loihi 2
Ecosystem." December 2, 2025.
https://uplatz.com/blog/the-convergence-of-spiking-neural-networks-and-neuromorphic-hardware-an-in-depth-analysis-of-intels-loihi-2-ecosystem/ -
ExoSwan Analysis. "Top Neuromorphic Computing Stocks 2026: Thinking on Microwatts." April 6, 2026.
https://exoswan.com/neuromorphic-computing-stocks/ -
Neuromorphica. "Aerospace and Defense Neuromorphic Chips with SDR." 2025.
https://neuromorphica.com/applications/aerospace-defense -
Conscium. "Major Neuromorphic Computing Projects." June 7, 2024.
https://conscium.com/explainers/major-neuromorphic-computing-projects/ -
Annett, Elise and Giordano, James. "Darwin Monkey: Next Generation
Neuromorphic Computing and Competition for Cognitive Capability and
Control." Digital Commons @ NDU, Institute for National Strategic Studies, 2025.
https://digitalcommons.ndu.edu/science-and-tech/ -
Giordano, James. "Strategic Innovation in the DoD FY 2026 RDTE
Budget: Leveraging Disruptive Technologies for Deterrence, Defense, and
Command and Control." Digital Commons @ NDU, 2025.
https://digitalcommons.ndu.edu/science-and-tech/
No comments:
Post a Comment